When diners
ask AI.

Roy Morgan's survey of 14,646 Australians found 58% of people aged 14 and over used an AI tool in an average four weeks in early 2026, and 45% used ChatGPT. Gemini reached 21%. That's a lot of people carrying a tool that will happily answer 'good Thai in Newtown that takes a booking for eight'.

For travel the habit is further along. Booking.com's 2025 AI Sentiment Report, with 1,013 Australians among more than 37,000 people surveyed, found half had used AI for some part of travel, and 37% to find restaurant recommendations while planning a trip. The same survey shows the limits: only 4% fully trusted AI, and 42% said they always fact-check what it tells them. So the AI answer starts the shortlist, and your Google profile and website still get checked before anyone books.

In the UK, a survey of more than 5,000 people for Zonal and NIQ found 19% already use AI tools to plan a visit to a pub, bar or restaurant. Among those people, 45% checked reviews, 43% average prices, 38% menus and 35% opening hours. I couldn't find an Australian figure that specific, so treat it as a guide to what people ask, not how many ask it here.

Google AI Overviews and AI Mode: Google's guide calls the method query fan-out, a set of related searches the model runs at the same time, then writes an answer from what it finds, with links. AI Mode also looks across reservation platforms and websites for bookable options, and Australia was one of eight markets that got that dining feature in April 2026.

Google Maps: Ask Maps, which puts Gemini inside Maps, began rolling out in Australia on 7 August 2026. It answers conversational questions about places and can compare hotel prices and availability, which matters if you run a motel in Port Macquarie as much as a café in Cronulla. At its US launch, it was reported to draw on more than 300 million places, including reviews from the Google community.

ChatGPT: OpenAI says ChatGPT can use your approximate location for local results such as restaurants, may show a map, and may show booking availability from OpenTable, which it lists as available globally. A site only appears in ChatGPT search if it lets OpenAI's crawler, OAI-SearchBot, in.

'Somewhere for a birthday dinner in Terrigal, Thai, good for a group of ten' is really five questions: who does Thai there, what does it cost, will they take ten, are they open Saturday, and what do people say about it. AI Mode runs searches like those and picks the venues where it finds answers. A venue that never wrote down that it takes groups of ten leaves the tool to guess, and it usually guesses someone else.

The test below builds that picture for your venue. Google doesn't publish the searches it runs, so the follow-ups are an illustration of the kind it runs. What matters is whether each fact exists, in text, somewhere a tool can read it.

Yext, a company that sells listings management, looked at 6.8 million citations from ChatGPT, Gemini and Perplexity and found 44% came from brands' own websites and 42% from listings, with food service drawing more on reviews and social (13.3%) than any other sector. Treat a vendor's study with some care, but it points the same way as Google's guide, which says Business Profiles can help businesses appear in AI responses.

For a venue, 'listings' means your Google profile, your booking platform, the delivery apps, Tripadvisor and the local guides that have written you up. If one says you close at 9 and another says 10, the tool has to pick one, or leave you out.

Google's guide, last updated 10 July 2026, says structured data isn't required for its AI features, and Google Search doesn't use llms.txt files. There's no ideal page length and no need to chop pages into chunks. Pages have to be indexed and eligible to show with a snippet. That's the whole technical list.

It also warns that making separate pages for every version of a question, to sway AI answers, breaks its scaled content abuse policy. One good menu page, one functions page and an about page that says plainly what you're good for will beat twenty thin pages called 'best brunch in' every suburb from Gosford to Newcastle.

None of this is new work. It's what a diner comparing three places on a Sunday night needs anyway. AI just makes the gaps obvious faster, because it fills them with someone else's version or skips you.

• The menu on a web page as text, with prices and dietary marks, matching the menu on your Google profile.
• Hours that are right today, with public holidays set ahead as special hours.
• A booking or ordering link on Google that works at 10pm, through a partner the tools can read.
• A plain line on what you're good for: groups, kids, dogs, late nights, quick lunches, a view.
• For accommodation: room types, what's included, check-in times and the cancellation policy, the same on your site and every booking site.
• Reviews that mention the dish or the occasion. Ask every guest the same way and let them write what they like.